Training machine learning models based on unlabeled data

Abstract

A method of labeling data and training a model is provided. The method includes obtaining a set of images. The set of images includes a first subset and a second subset. The first subset is associated with a first set of labels. The method also includes generating a set of pseudo labels for the set of images and a second set of labels for the second subset based on the first subset, the second subset, a first machine learning model, and a domain adaption model. The method further includes generating second machine learning model. The second machine learning model is generated based on the set of images, the set of pseudo labels, the first set of labels, and the second set of labels. The second set of labels is updated based on one or more inferences generated by the second machine learning model.

What is claimed is: 
     
       1. A method, comprising:
 obtaining a set of images, wherein:
 the set of images comprises a first subset and a second subset; and 
 the first subset is associated with a first set of reference labels; 
 
 generating a set of pseudo labels for the set of images and a second set of labels for the second subset based on the first subset, the second subset, a first machine learning model, and a domain adaption model including a source domain and a target domain, wherein the source domain and the target domain are based on clusters of features determined by the first machine learning model, such that a radius around a centroid of each of the clusters of features separates the source domain and the target domain; 
 generating second machine learning model, wherein:
 the second machine learning model is generated based on the set of images, the set of pseudo labels, the first set of reference labels, and the second set of labels; and 
 the second set of labels is updated based on one or more inferences generated by the second machine learning model; and 
 training a third machine learning model based on the set of images and the second set of labels. 
 
 
     
     
       2. The method of  claim 1 , wherein generating the set of pseudo labels and the second set of labels comprises:
 determining a set of clusters of features based on the first machine learning model and set of images; and 
 associating each cluster of the set of clusters with an initial pseudo label to determine an initial set of pseudo labels. 
 
     
     
       3. The method of  claim 2 , wherein generating the set of pseudo labels and the second set of labels further comprises:
 generating the domain adaption model based on the initial set of pseudo labels; and 
 generating the set of pseudo labels based on the set of images and the domain adaption model. 
 
     
     
       4. The method of  claim 3 , wherein generating the set of pseudo labels and the second set of labels further comprises:
 determining a set of majority initial pseudo labels for the set of reference labels, wherein each majority initial pseudo label of the set of majority initial pseudo labels is associated with a reference label from the set of reference labels. 
 
     
     
       5. The method of  claim 4 , wherein generating the set of pseudo labels and the second set of labels further comprises:
 generating the second set of labels based on the set of majority initial pseudo labels for each real label. 
 
     
     
       6. The method of  claim 1 , wherein generating second machine learning model comprises:
 generating a set of inferences based on the set of data and the second machine learning model; and 
 updating one or more of pseudo labels of the set of pseudo labels based on the set of inferences. 
 
     
     
       7. The method of  claim 6 , wherein updating the one or more of pseudo labels of the set of pseudo labels based on the set of inferences comprises:
 determining one or more inferences associated with one or more confidence levels that are greater than a threshold confidence; 
 determining the one or more pseudo labels based on the one or more inferences and a majority pseudo label for each real label; and 
 updating the one or more pseudo labels based on the majority pseudo label. 
 
     
     
       8. A system comprising:
 a memory to store data; and 
 a processing device, operatively coupled to the memory, to:
 obtain a set of images, wherein:
 the set of images comprises a first subset and a second subset; and 
 the first subset is associated with a first set of reference labels; 
 
 generate a set of pseudo labels for the set of images and a second set of labels for the second subset based on the first subset, the second subset, a first machine learning model, and a domain adaption model including a source domain and a target domain, wherein the source domain and the target domain are based on clusters of features determined by the first machine learning model, such that a radius around a centroid of each of the clusters of features separates the source domain and the target domain; 
 generate second machine learning model, wherein:
 the second machine learning model is generated based on the set of images, the set of pseudo labels, the first set of reference labels, and the second set of labels; and 
 the second set of labels is updated based on one or more inferences generated by the second machine learning model. 
 
 
 
     
     
       9. The system of  claim 8 , wherein to generate the set of pseudo labels and the second set of labels the processing device is to:
 determine a set of clusters of features based on the first machine learning model and set of images; and 
 associate each cluster of the set of clusters with an initial pseudo label from an initial set of pseudo labels. 
 
     
     
       10. The system of  claim 9 , wherein generating the set of pseudo labels and the second set of labels further comprises:
 generating the domain adaption model based on the initial set of pseudo labels; and 
 generating the set of pseudo labels based on the set of images and the domain adaption model. 
 
     
     
       11. The system of  claim 10 , wherein generating the set of pseudo labels and the second set of labels further comprises:
 determining a set of majority initial pseudo labels for the set of reference labels, wherein each majority initial pseudo label of the set of majority initial pseudo labels is associated with a reference label from the set of reference labels. 
 
     
     
       12. The system of  claim 11 , wherein generating the set of pseudo labels and the second set of labels further comprises:
 generating the second set of labels based on the set of majority initial pseudo labels. 
 
     
     
       13. The system of  claim 8 , wherein generating second machine learning model comprises:
 generating a set of inferences based on the set of data and the second machine learning model; and 
 updating one or more of pseudo labels of the set of pseudo labels based on the set of inferences. 
 
     
     
       14. The system of  claim 13 , wherein updating the one or more of pseudo labels of the set of pseudo labels based on the set of inferences comprises:
 determining one or more inferences associated with one or more confidence levels that are greater than a threshold confidence; 
 determining the one or more pseudo labels based on the one or more inferences and a majority pseudo label; and 
 updating the one or more pseudo labels based on the majority pseudo label. 
 
     
     
       15. A non-transitory computer-readable storage medium having instructions stored thereon that, when executed by a processing device, cause the processing device to:
 obtain a set of images, wherein:
 the set of images comprises a first subset and a second subset; and 
 the first subset is associated with a first set of reference labels; 
 
 generate a set of pseudo labels for the set of images and a second set of labels for the second subset based on the first subset, the second subset, a first machine learning model, and a domain adaption model including a source domain and a target domain, wherein the source domain and the target domain are based on clusters of features determined by the first machine learning model, such that a radius around a centroid of each of the clusters of features separates the source domain and the target domain; 
 generate second machine learning model, wherein:
 the second machine learning model is generated based on the set of images, the set of pseudo labels, the first set of reference labels, and the second set of labels; and 
 the second set of labels is updated based on one or more inferences generated by the second machine learning model. 
 
 
     
     
       16. The non-transitory computer-readable storage medium of  claim 15 , wherein generating the set of pseudo labels and the second set of labels comprises:
 determining a set of clusters of features based on the first machine learning model and set of images; and 
 associating each cluster of the set of clusters with an initial pseudo label from an initial set of pseudo labels. 
 
     
     
       17. The non-transitory computer-readable storage medium of  claim 16 , wherein generating the set of pseudo labels and the second set of labels further comprises:
 generating the domain adaption model based on the initial set of pseudo labels; and 
 generating the set of pseudo labels based on the set of images and the domain adaption model. 
 
     
     
       18. The non-transitory computer-readable storage medium of  claim 17 , wherein generating the set of pseudo labels and the second set of labels further comprises:
 determining a set of majority initial pseudo labels for the set of reference labels, wherein each majority initial pseudo label of the set of majority initial pseudo labels is associated with a reference label from the set of reference labels. 
 
     
     
       19. The non-transitory computer-readable storage medium of  claim 18 , wherein generating the set of pseudo labels and the second set of labels further comprises:
 generating the second set of labels based on the set of majority initial pseudo labels. 
 
     
     
       20. The non-transitory computer-readable storage medium of  claim 15 , wherein generating second machine learning model comprises:
 generating a set of inferences based on the set of data and the second machine learning model; and 
 updating one or more of pseudo labels of the set of pseudo labels based on the set of inferences.